SUITABILITY OF POCL3 DIFFUSION PROCESSES WITH IN-SITU OXIDATION FOR FORMING LASER-DOPED SELECTIVE EMITTERS WITH LOW CARRIER RECOMBINATION S. Werner, E. Lohmüller, J. Weber, A. Wolf Fraunhofer Institute for Solar Energy Systems ISE, Heidenhofstr. 2, 7911 Freiburg, Germany Phone: +49 761-4588 549; e-mail: sabrina.werner@ise.fraunhofer.de ABSTRACT: Laser-doped selective emitters feature the advantage of more effective shielding of minority charge carriers from the metal contacts while allowing for low emitter saturation current density je in the photoactive area. The formation of emitters by diffusion processes using phosphorus oxychloride (POCl3) with incorporated oxidation gains more and more attention as it allows for low je below 9 fa/cm² on textured surface with silicon nitride passivation in industrial cycle times. Hence, the combination of both POCl3 diffusion with oxidation and laser doping is very interesting. We examine four different POCl3 diffusions with oxidation and one reference POCl3 diffusion without oxidation in terms of their suitability for selective emitter laser doping. Detailed characterizations of the grown layers on the silicon surface are performed after diffusion with respect to the individual layer nesses of the phosphosilicate glass (PSG) and the intermediate silicon dioxide (SiO2) layer as well as the stacks total phosphorus doses. The as-diffused depth-dependent charge carrier concentration profiles show that a second PSG deposition step attached after drive-in to the diffusion processes hardly impacts their course. We find that POCl3 diffusions with oxidation especially those with second deposition step allow for effective laser doping. Thereby, the intermediate SiO2 layer ness plays a key role: the er the layer is the less phosphorus can be incorporated additionally from the PSG layer into the silicon. Keywords: oxidation, POCl3 diffusion, selective emitter, laser doping, phosphosilicate glass 1 INTRODUCTION Tube furnace diffusion processes using phosphorus oxychloride (POCl3) as liquid dopant precursor are the dominating emitter formation technology for p-type silicon solar cells [1]. Improving POCl3 diffusion processes and the resulting emitter doping profiles is essential for further increasing the energy conversion efficiency of silicon solar cells. The reduction of emitter recombination, enabled by a fast progress in front contact screen printing paste properties, has been of major interest in the past few years. An approach to realize less emitter recombination is the reduction of the surface doping concentration by implementing an oxidation into the POCl3 diffusion process [2 9]. When applied to passivated emitter and rear solar cells (PERC) [1], energy conversion efficiencies beyond 21% are reported [6,11,12]. As the charge carrier recombination at the metal contacts is high for that kind of doping profiles with low surface doping concentration [13], one option for its decrease is the selective emitter approach. Recently, selective emitters came back into focus for PERC solar cells targeting the 22% efficiency regime and above [6,11,12,14,15]. The selective emitter can be formed by, e.g., local laser doping out of the phosphosilicate glass (PSG) layer [16]. Correctly, POCl3 diffusions form a stack layer consisting of a PSG layer and a silicon dioxide (SiO2) layer: the SiO2 layer separates the PSG layer from the silicon surface [9,8,17,18]. Hereby, the SiO2 layer features a much lower phosphorus concentration than the phosphorus-rich PSG layer. For laser doping, sufficiently high phosphorus content wi this PSG/SiO2 stack layer is necessary. Hence, this paper investigates the combination of tube furnace POCl3 diffusion with oxidation and the formation of selective emitters by laser doping out of the grown PSG/SiO2 stack layer. 2 APPROACH The approach pursued in this paper to ensure sufficient phosphorus content in the PSG/SiO2 stack layer for effective laser doping is the implementation of a second deposition step with active nitrogen (N2) gas flow through the POCl3 bubbler (i.e. active N2-POCl3 flow) at the end of the diffusion process. We investigate two pairs of POCl3 diffusion processes that feature an oxidation and that are performed with and without a second deposition step, respectively. POCl3 diffusion without oxidation but second deposition step serves as reference. Table I summarizes the examined POCl3 diffusion processes. Process details and the labeling of the processes are discussed in detail in section 2.1. The schematic experiment plan in Fig. 1 depicts the three groups of test samples fabricated and the different characterization approaches applied in this work. Fullsquare p-type Czochralski-grown silicon (Cz-Si) wafers with an edge length of 156 mm serve as starting material. After either saw damage etching or alkaline texturing, the five different tube furnace POCl3 diffusion processes form the emitter and the PSG/SiO2 stack layer. The different POCl3 diffusion processes, the applied Table I: Summary of the five investigated POCl3 diffusion processes. The emitter dark saturation current densities je, determined on alkaline textured surface after passivation by a conventional SiNx layer and firing, are taken from Refs. [4,11,13]. (n.d.: not determined) POCl3 process 1 st dep. drive-in (N2-POCl3) (O2 share). je (fa/ cm²) Ref med < 4% yes 42 low 1% - 73 low 1% yes n.d. high 1% - 85 high 1% yes n.d.
Group 1 Group 2 Group 3 p-typ Cz-Si, 156 mm edge length Saw damage etch Alkaline texturing POCl 3 diffusion variation Wafer cutting PSG/SiO 2 etch Reflection meas. Ellipsometry ICP-OES Laser doping Selective etching Four point probe measurements Ellipsometry ECV Figure 1: Experiment plan for investigating laser doping out of the SiO2/PSG stack layer formed by different POCl3 diffusion processes. (ECV: electrochemical capacitance-voltage measurements; ICP-OES: inductively coupled plasma optical emission spectrometry). further process steps, and the utilized characterization methods are being discussed in the following sections. 2.1 POCl3 diffusion processes We target at low content of inactive phosphorus at the surface in the as-diffused emitter and thus, low charge carrier recombination in the homogeneously-doped area. For this purpose, we select the five POCl3 diffusion processes summarized in Table I. For three of the five processes, we previously characterized their emitter dark saturation current density je in Refs. [4,11,13]. The je values on alkaline textured surface passivated by a conventional silicon nitride (SiNx) passivation layer that is formed by plasma-enhanced chemical vapor deposition are found to be between 42 fa/cm² je 85 fa/cm² after simulated contact firing (see also Table I, rightmost column). Table I also contains characteristic process parameters for the examined POCl3 diffusion processes. We label the processes with respect to the used oxidation and the ness of the resulting total PSG/SiO2 stack layer. The reference process Ref has no oxidation as the oxygen (O2) share in the gas atmosphere during drive-in is below 4% but features a second PSG deposition with active N2-POCl3 flow after the drive-in. For the other four processes, the drive-in with 1% O2 share is similar. Corresponding to the name convention, the processes and feature either a or a PSG/SiO2 stack layer after diffusion, respectively (see section 3.1). The N2-POCl3 and O2 gas flows during the first deposition are a factor two to three lower for the process than for the process. The suffix indicates that a second PSG deposition step with active N2-POCl3 flow follows the drive-in step. The two process variations with the same second deposition step as used for the reference process are accordingly named and. 2.2 Determination of layer nesses by means of selective etching In order to extract the individual nesses of the PSG and SiO2 layers formed during the five POCl3 diffusions, the wafers with saw damage etched surface of group 1 in Fig. 1 are cut into 36 pieces each after diffusion. Ellipsometry measurements are performed before and after stepwise etching the stack layers in.1wt% hydrofluoric acid (HF) solution [9]. As the etching rate is dependent on the phosphorus concentration wi the respective layer [8,9], the ness of each layer can be extracted. The detailed procedure is described in Ref. [9]. Applying this procedure, we examine the individual nesses of the PSG and SiO2 layers of the five diffusion processes and the impact of a second deposition step on these layer nesses for diffusion processes with insitu oxidation. 2.3 Determination of phosphorus doses and charge carrier concentration profiles We use the inductively coupled plasma optical emission spectrometry (ICP-OES) [19] to determine the phosphorus dose wi the total PSG/SiO2 stack layer on alkaline textured surfaces. Therefore, the stack layers of the wafers of group 2 in Fig. 1 are etched in diluted HF and the solution is analyzed with respect to its phosphorus concentration. After the PSG/SiO2 stack layer has been removed, the samples are further characterized with respect to their doping properties. Electrochemical capacitance-voltage measurements (ECV) [2] yield the charge carrier concentration profiles. The profiles are scaled to match the sheet resistances Rsh [13,21], which are locally determined by four point probe (4pp) measurements around the ECV spots. 2.4 Suitability for laser doping To examine the suitability of the POCl3 diffusion processes with respect to laser doping, we use a pulsed ultraviolet (UV) ns-laser with a wavelength λlaser = 355 nm [22]. Subsequent to reflection measurements of the PSG/SiO2 stack layer on textured surface for the wafers of group 3 in Fig. 1, we prepare 2 x 2 cm²-large test fields as shown in Fig. 2 applying four different pulse energies Ep,set = {62 µj; 74 µj; 9 µj; 12 µj} at constant pulse-to-pulse and line-to-line distances dpulse = 15 µm and dline = 25 µm, respectively. Then, the evolution of Rsh for the applied laser processes for the different POCl3 diffusions is determined by 4pp measurements. 3 RESULTS 3.1 PSG/SiO2 layer nesses after POCl3 diffusion Fig. 3 shows the individual nesses of the PSG and SiO2 layers dpsg and dsio2, respectively, for the five diffusion processes investigated in this work. For the reference process Ref, the PSG layer with dpsg = (24 ± 2) nm is about four times er than the SiO2 layer with dsio2 = (6 ± 1) nm. Chen et al. [18] reported a nearly constant SiO2 layer ness dsio2 6 nm for different deposition times and POCl3 flow rates for common diffusion processes without oxidation, while they found the total stack layer nesses d Figure 2: Image scan of an alkaline textured Cz-Si wafer with PSG/SiO2 stack layer after laser doping. The visible square fields with a size of 2 x 2 cm² were processed with different laser parameters.
Thickness d (nm) 8 6 4 2 d PSG 24.3 6. d SiO2 Ref 9.8 23. 19.1 11.7 23.1 26.5 49.5 5.1 Figure 3: Individual layer nesses dpsg and dsio2 of the PSG and the SiO2 layer, respectively, formed on the silicon surface during the five POCl3 tube furnace diffusions from Table I. The individual nesses are extracted by the selective etching procedure described in Ref. [9]. to be between 1 nm < d < 4 nm. Hence, the layer nesses found for our reference process without oxidation Ref are in accordance with those found by Chen et al., although Chen et al. investigated processes without second PSG deposition. In contrast, dsio2 for the process in Fig. 3 is significantly larger with dsio2 = (19 ± 2) nm. This er intermediate SiO2 layer is formed by the insitu oxidation [9]. During oxidation, the oxygen from the atmosphere diffuses through the PSG and SiO2 layers and reacts with the silicon at the surface to form SiO2. Attaching a second PSG deposition for process insitu, the second deposition clearly leads to an increase in dpsg to dpsg = (23 ± 1) nm, while dsio2 is reduced to dsio2 = (12 ± 2) nm. This result (i.e. the increase in dpsg while dsio2 is decreased) is in accordance with results we found in Ref. [9], where we investigated the influence of a second PSG deposition step on the PSG/SiO2 stack layer for a diffusion process without insitu oxidation. We propose that the reduction in dsio2 results from the diffusion of phosphorus from the PSG layer to the PSG/SiO2 interface, where they react with SiO2 to form additional PSG [18]. The PSG layer ness for process is comparable to that of the reference process. Processes and feature significantly higher total layer nesses d than the other processes as the N2-POCl3 and O2 gas flows during first PSG deposition are two to three times higher; see Table I. The strong first PSG deposition for results in a total ness of d 7 nm, while the weak first PSG deposition for the reference process and both -processes results in d 3 nm. For diffusion, dsio2 = (5 ± 2) nm is more than twice as large as dpsg = (23 ± 2) nm. The attached second PSG deposition for process hardly affects these layer nesses: dsio2 = (5 ± 2) nm and dpsg = (27 ± 2) nm. Only dpsg is about 3 nm larger. In contrast to the process, dsio2 is not decreased for process by adding the second PSG deposition step. As the intermediate SiO2 layer is more than twice as for Phosphorus dose D Phos (1 15 cm -2 ) 3 2 1 Ref no Figure 4: Total phosphorus dose wi the PSG/SiO2 stack layer for POCl3 diffusion processes with and without second deposition step, determined by ICP-OES., it seems that there is either a critical SiO2 layer ness, which represses the increase of dpsg by a second deposition step, or a critical PSG layer ness, from which point the PSG layer growth occurs slower. Also remarkable is the fact that for all processes that feature the (same) second deposition step, similar dpsg values between 23 nm dpsg < 27 nm are found, irrespective on the gas flows used during first deposition and drive-in. On the other hand, dsio2 varies significantly between 6 nm dsio2 < 5 nm. 3.2 Phosphorus dose of the PSG/SiO2 stack layers The phosphorus dose wi the total PSG/SiO2 stack layers on textured surface is determined by ICP-OES. The ICP-OES measurement gives the phosphorus weight wi the examined solution in µg/l. This quantitiy is converted to the amount of phosphorus atoms and then divided by the sample area. As the diffusion occurs on both wafer surfaces, that value is further divided by two to obtain the phosphorus dose DPhos for one wafer side. Fig. 4 shows the obtained phosphorus doses DPhos for the five diffusion processes. It is evident that the second PSG deposition step leads to significantly higher DPhos that ranges between (2.3 ±.3) 1 15 cm -2 DPhos (2.8 ±.4) 1 15 cm -2. For process, the second deposition leads to an increase in DPhos from DPhos (.61 ±.9) 1 15 cm -2 to DPhos (2.3 ±.3) 1 15 cm -2, which corresponds to a factor of four. Process with very intermediate SiO2 layer, see Fig. 3, shows a slighter increase in DPhos due to the second deposition step. This is consistent with the assumption that the majority of the phosphorus is located in the PSG layer [9,18] and hence a slight increase in dpsg correlates with a slightly higher DPhos. 3.3 Phosphorus doping profiles after POCl3 diffusion Fig. 5(a) shows the as-diffused charge carrier concentration profiles of the reference process and the two diffusion processes and, determined by ECV measurements on textured surface. The profile depth dprof at a dopant concentration N = 1 17 cm -3 is between 32 nm < dprof < 4 nm. It is clear that the surface charge carrier concentration Nsurf = (6.5 ±.7) 1 19 cm -3 for the reference process Ref is lower than for the -process with Nsurf = (2.1 ±.2) 1 2 cm -3. On the other hand, the course of the profiles for processes
Charge carrier concentration N (cm -3 ) 1 2 1 19 1 18 Figure 5: Charge carrier concentration profiles obtained by the ECV technique after PSG etching on textured surface for the five diffusion processes stated in (a) and (b). The emitter sheet resistances Rsh as well as the integrated total charge carrier dose QECV are stated. and is very similar. Hence, the performed second deposition step hardly affects the resulting doping profile. This is confirmed by the integrated total charge carrier dose QECV given in Fig. 5(a), which is similar for both processes wi the measurement uncertainty. Although the PSG/SiO2 layer nesses are very similar for the processes Ref and (Fig. 3), the courses of the profiles are different. Fig. 5(b) shows the charge carrier concentration profiles of, the reference process and the two diffusion processes and. Here, dprof is very similar for all processes with dprof 4 nm. Both -processes with Nsurf = (1.8 ±.2) 1 2 cm -3 result in very similar charge carrier concentration profiles as obtained for the two -processes. Again, it is clear that the performed second deposition step hardly affects the resulting doping profile: the charge carrier concentrations are very similar either without or with second deposition. This is also confirmed by the integrated total charge carrier dose QECV given in Fig. 5(b), which is similar for both processes. R sh Q ECV ( /sq) (1 14 cm -2 ) Ref 15 5.5 96 7.4 96 7. 1 17..1.2.3.4 (a) Depth d (μm) Charge carrier concentration N (cm -3 ) 1 2 1 19 1 18 1 17..1.2.3.4 (b) Depth d (μm) R sh Q ECV ( /sq) (1 14 cm -2 ) Ref 15 5.5 89 7.8 86 7.9 Reflection R (%) 4 3 2 1 Ref 3 35 4 45 3 5 7 9 11 Wavelength (nm) Figure 6: Reflection measurements on textured surface with the PSG/SiO2 stack layers from the indicated POCl3 diffusion processes. 3.4 Impact of laser doping on sheet resistance As the laser doping process is at least partly dependent on the coupling of the laser radiation into the silicon material, the absorption A of the samples is of interest. In order to obtain A, we measure the reflection R which is dependent on the surface morphology and the total layer ness of the PSG/SiO2 stack. Commonly, lasers with wavelengths between 355 nm λlaser 13 nm are used for laser doping approaches [23]. Wi this work, we use a UV ns-laser at λlaser = 355 nm. Fig. 6 shows R as a function of wavelength λ. R is significantly lower at λlaser = 355 nm for both diffusion processes compared to the other processes as they feature a er PSG/SiO2 stack layer. As the transmission is zero for λlaser, the absorption is A = 1 R. For the results plotted in Fig. 7, we multiplied the applied pulse energy EP,set with the corresponding A, in order to account for the differences in reflection. However, note that the reflectance changes dramatically during the application of each laser pulse due to local heating and melting of the wafer surface. Fig. 7 summarizes the obtained sheet resistances Rsh,4pp, determined by 4pp measurements after laser doping. All three images (a) to (c) have in common that Rsh,4pp decreases continuously with increasing EP. For the reference POCl3 diffusion process with second deposition Ref 2 nd (no oxidation) in Fig. 7(a), Rsh,4pp is reduced from Rsh,4pp 1 Ω/sq to Rsh,4pp 2 Ω/sq if EP is increased to EP 1 µj. For both diffusion processes with oxidation and second deposition in Fig. 7(a), also a significant reduction in Rsh,4pp to Rsh,4pp 25 Ω/sq and Rsh,4pp 3 Ω/sq is seen for the processes and -, respectively. However, the reductions in Rsh,4pp are somewhat weaker pronounced than for the reference. Referring to Figs. 3 and 4, it is clear that the different nesses of the intermediate SiO2 layer is the dominating property which causes the differences in the obtained Rsh,4pp values after laser doping in Fig. 7(a), as all three POCl3 diffusion processes feature comparable dpsg and DPhos. The er the intermediate SiO2 layer is, the less is the achievable reduction in Rsh,4pp. The presence of a intermediate SiO2 layer (dsio2 5 nm) thus retards the in-corporation of phosphorus from the PSG layer into the silicon surface considerably compared to the ner SiO2 layer (dsio2 12 nm) present for the 3 2 1 355 nm
Emitter sheet resistance R sh,4pp ( /sq) 1 8 6 5 4 3 2 1 decrease in d SiO2 Ref - 4 5 6 7 8 9 111-4 5 6 7 8 9 111-4 5 6 7 8 9 111 Pulse energy E P (μj) Pulse energy E P (μj) Pulse energy E P (μj) (a) (b) (c) Figure 7: Emitter sheet resistances Rsh,4pp after laser doping for the pulse energies EP which correspond to the applied pulse energies EP,set times the absorption A of the respectively indicated diffusion processes in (a) to (c). The pulse-to-pulse as well as the line-to-line distance is constant for all EP. One hundred 4pp measurements are performed for all diffusion processes without laser doping (-), while ten 4pp measurements are performed per EP and per diffusion process. diffusion process. The even ner intermediate SiO2 layer (dsio2 6 nm) of the reference process allows for an even stronger laser doping. Fig. 7(b) shows the impact of the second deposition step during POCl3 diffusion for the processes. Obviously, the second deposition and thus, the higher DPhos as shown in Fig. 4, allows for significantly higher doping resulting in lower Rsh,4pp values. Remember that the second deposition step not only increases DPhos but also leads to the decrease in the intermediate SiO2 layer ness; see Fig. 3. A ner SiO2 layer also allows for higher laser doping, compare Fig. 7(a). For the processes in Fig. 7(c), Rsh,4pp decreases with increasing EP, but only slightly lower values are seen for the POCl3 diffusion process with second deposition at each EP. This can be explained by the only slight increase of the phosphorous dose due to the second deposition (see Fig. 4). The intermediate SiO2 layer (see Fig. 3) seems to prevent the in-diffusion of phosphorus atoms from the PSG into the silicon for both processes, although slightly more phosphorus is provided wi the PSG/SiO2 stack layer for (see Fig. 4). 5 SUMMARY AND CONCLUSION Five different diffusion processes utilizing phosphorus oxychloride (POCl3) as liquid dopant precursor are examined with respect to their as-diffused properties and their suitability with laser doping from phosphosilicate glass (PSG) to form a selective emitter structure. Therefore, variations in deposition parameters, oxidation (e.g. high oxygen flow during drive-in), and the influence of a second PSG deposition step (e.g. active N2 flow through the POCl3 bubbler after drive-in) are investigated. The POCl3 diffusions with oxidation result in grown stack layers consisting of silicon dioxide (SiO2) and PSG, whose intermediate SiO2 layers with a ness dsio2 = (19 ± 2) nm or dsio2 = (5 ± 2) nm are twice as as the PSG layers with nesses dpsg = (1 ± 1) nm and dpsg = (23 ± 1) nm, respectively. The total PSG/SiO2 stack layer ness depends on the chosen gas flows during deposition. The higher POCl3 and O2 gas flows result in er stack layers. The implementation of a second PSG deposition step for the diffusion process with oxidation results in increased PSG layer nesses, while the charge carrier concentration profiles hardly change. The SiO2 layer ness decreases if the intermediate SiO2 layer before second deposition does not exceed a critical ness. Or, there is a critical PSG layer ness from which the PSG layer growth occurs slower. The reference POCl3 diffusion without oxidation but with second deposition features a SiO2 with dsio2 = (6 ± 2) nm, while the PSG layer is about four times er. We find that the POCl3 diffusion processes with second PSG deposition feature a higher total phosphorus dose DPhos wi the PSG/SiO2 stack layers. DPhos ranges between (2.3 ±.3) 1 15 cm -2 DPhos (2.8 ±.4) 1 15 cm -2 for all processes with second deposition. Laser doping applied to the PSG/SiO2 stack layers reduces the emitter sheet resistance Rsh with increasing laser pulse energy. It is found that for the POCl3 diffusion processes with second PSG deposition, Rsh can be significantly more reduced compared with the POCl3 diffusion processes without second deposition as long as the intermediate SiO2 layer ness is a few 1 nm or less. From Rsh 9 Ω/sq after diffusion, Rsh can be reduced to 15 Ω/sq < Rsh < 35 Ω/sq by laser doping. This result suggests that the additionally provided phosphorus from the second deposition step is advantageous for laser doping. For POCl3 diffusion processes with oxidation, the reduction in Rsh by the laser process is weaker for er intermediate SiO2 layers. For SiO2 layer nesses above about 5 nm, laser doping results in similar Rsh values measured on samples either diffused with or without second PSG deposition step. In summary, our results show that POCl3 diffusion processes with oxidation are suitable for the formation of selective emitters by laser doping. ACKNOWLEDGEMENTS The authors would like to thank all colleagues at the Fraunhofer ISE PV-TEC; especially C. Harmel for laser
processing, A. Fischer and S. Maus for selective etching, A. Moldovan for ICP-OES measurements, S. Schmidt and S. Dholipet Nagendra Kumar for ECV measurements and M. Berlin for 4pp measurements. This work was funded by the German Federal Ministry for Economic Affairs and Energy wi the research project POLDI under contract number 32479D. REFERENCES [1] ITRPV, 217, "International Technology Roadmap for Photovoltaic (ITRPV) - Results 216", available onl.: http://www.itrpv.net/reports/ Downloads/217/. [2] A. Cuevas, "A good recipe to make silicon solar cells", Proc. 22nd IEEE PVSC, Las Vegas, USA, 1991, pp. 466 47. [3] A. Kimmerle, R. Woehl, A. Wolf et al., "Simplified front surface field formation for back contacted silicon solar cells", Energy Proced., vol. 38, pp. 278 282, 213. [4] S. Werner, E. Lohmüller, S. Maier et al., "Process optimization for the front side of p-type silicon solar cells", Proc. 29th EU PVSEC, Amsterdam, The Netherlands, 214, pp. 1342 1347. [5] S. Werner, E. Lohmüller, A. Wolf et al., "Extending the limits of screen-printed metallization of phosphorus- and boron-doped surfaces", Sol. Energy Mater. Sol. Cells, vol. 158, pp. 37 42, 216. [6] T. Dullweber, H. Hannebauer, S. Dorn et al., "Emitter saturation currents of 22 fa/cm 2 applied to industrial PERC cells approaching 22% conversion efficiency", Proc. 32nd EU PVSEC, Munich, Germany, 216, pp. 335 339. [7] A. Wolf, A. Kimmerle, S. Werner et al., "Status and perspective of emitter formation by POCl3- diffusion", Proc. 31st EU PVSEC, Hamburg, Germany, 215, pp. 414 419. [8] P. A. Basore, J. M. Gee, M. E. Buck et al., "Simplified high-efficiency silicon cell processing", Proc. 7th Int. PVSEC, Nagoya, Japan, 1993, pp. 61 64. [9] S. Werner, S. Mourad, W. Hasan et al., "Structure and composition of phosphosilicate glass systems formed by POCl3 diffusion", Energy Proced., in press, 217. [1] A. W. Blakers, A. Wang, A. M. Milne et al., "22.8% efficient silicon solar cell", Appl. Phys. Lett., vol. 55, no. 13, pp. 1363 1365, 1989. [11] S. Werner, E. Lohmüller, P. Saint-Cast et al., "Key aspects for fabrication of p-type Cz-Si PERC solar cells exceeding 22% conversion efficiency", Proc. 33rd EU PVSEC, Amsterdam, The Netherlands, 217. [12] P. Saint-Cast, S. Werner, J. Greulich et al., "Analysis of the losses of industrial-type PERC solar cells", Phys. Status Solidi A, 216. [13] S. Werner, E. Lohmüller, S. Maier et al., "Challenges for lowly-doped phosphorus emitters in silicon solar cells with screen-printed silver contacts", Energy Proced., in press, 217. [14] G. Fischer, M. Müller, S. Steckemetz et al., "Model based continuous improvement of industrial p-type PERC technology beyond 21% efficiency", Energy Proced., vol. 77, pp. 515 519, 215. [15] M. Müller, B. Bitnar, G. Fischer et al., "Loss analysis of 22% efficient industrial PERC solar cells", Energy Proced., in press, 217. [16] L. Ventura, A. Slaoui, J. C. Muller, "Realization of selective emitters by rapid thermal and laser assisted techniques", Proc. 13th EU PVSEC, Nice, France, 1995, pp. 1578 1581. [17] A. Cuevas, P. A. Basore, G. Giroult Matlakowski et al., "Surface recombination velocity of highly doped n type silicon", J. Appl. Phys., vol. 8, no. 6, pp. 337 3375, 1996. [18] R. Chen, H. Wagner, A. Dastgheib-Shirazi et al., "A model for phosphosilicate glass deposition via POCl3 for control of phosphorus dose in Si", J. Appl. Phys., vol. 112, no. 12, p. 124912, 212. [19] Nölte, Joachim, "ICP - Emissionsspektrometrie für PraktikerWiley, 22. [2] E. Peiner, A. Schlachetzki, D. Krüger, "Doping profile analysis in Si by electrochemical capacitance-voltage measurements", J. Electrochem. Soc., vol. 142, no. 2, pp. 576 58, 1995. [21] R. Bock, P. P. Altermatt, J. Schmidt, "Accurate extraction of doping profiles from electrochemical capacitance voltage measurements", Proc. 23rd EU PVSEC, Valencia, Spain, 28, pp. 151 1513. [22] U. Jäger, S. Mack, C. Wufka et al., "Benefit of selective emitters for p-type silicon solar cells with passivated surfaces", IEEE J. Photovoltaics, vol. 3, no. 2, pp. 621 627, 213. [23] U. Jäger, M. Okanovic, M. Hörteis et al., "Selective emitter by laser doping from phosphosilicate glass", Proc. 24th EU PVSEC, Hamburg, Germany, 29, pp. 174 1743.